Meeting Banner
Abstract #0188

Multicentre Prospective Classification of Childhood Brain Tumours Based on Metabolite Profiles Derived from 1H MRS

Nigel Paul Davies1,2, Simrandip Gill2,3, Theodoros N. Arvanitis3,4, Dorothee Auer5, Richard Grundy6,7, Franklyn A. Howe8, Darren Hargrave9, Tim Jaspan7, Lesley MacPherson3, Kal Natarajan1,3, Geoffrey Payne9,10, Dawn Saunders11, Yu Sun2,3, Martin Wilson2,3, Andrew C. Peet2,3

1Medical Physics, University Hospitals Birmingham NHS Foundation Trust, Birmingham, United Kingdom; 2Cancer Sciences, University of Birmingham, Birmingham, United Kingdom; 3Birmingham Children's Hospital NHS Foundation Trust, Birmingham, United Kingdom; 4Department of Electrical, Electronic & Computer Engineering, University of Birmingham, Birmingham, United Kingdom; 5Academic Radiology, University of Nottingham, Nottingham, United Kingdom; 6Children's Brain Tumour Research Centre, University of Nottingham, Nottingham, United Kingdom; 7University Hospital Nottingham, Nottingham, United Kingdom; 8St. George's University of London, London, United Kingdom; 9Royal Marsden Hospital, London, United Kingdom; 10Institute of Cancer Research, London, United Kingdom; 11Great Ormond Street Hospital, London, England, United Kingdom


1H MRS provides non-invasive metabolite profiles of brain tumours aiding diagnosis and potentially improving characterisation. In this study we perform a large prospective multicentre evaluation of 1H MRS as a diagnostic tool for grading childhood brain tumours. The classifier was trained on metabolite profiles derived using TARQUIN from 123 single-voxel MRS acquired using a standard protocol on two 1.5T scanners in a single centre. Testing was performed using 110 cases acquired prospectively across 4 different centres with some variations in echo time and field strength. The overall classification accuracy for identifying high grade versus low grade tumours was 86%.